A novel clustering-based filter for impulsive noise reduction in electromagnetic tomography (EMT)

Demodulation is a crucial step in signal processing for electromagnetic tomography (EMT) systems. Impulse noise appears in EMT, in particular, when excitation current is switched from one coil to another. This is mainly a result of the fact that the discontinuity of current creates voltage spikes due to back electromotive force. This paper proposes a novel clustering-based method for reducing this impulsive noise on the basis of Kalman filter demodulation. The Kalman filter is capable of demodulating the multiplexing signal reclusively, but it introduces dynamic oscillations at the same time. The proposed clustering method is able to successfully separate the signal from the impulse noise and smoothen the dynamic oscillations after demodulation. Simulation and experimental results show that the proposed clustering method can improve the signal-to-noise ratio by more than 20 dB in both the real and imaginary parts of EMT measurements. This technique is also useful in multi-channel coil systems where coil switching is necessary.